|Publication number||US7937653 B2|
|Application number||US 11/032,817|
|Publication date||May 3, 2011|
|Filing date||Jan 10, 2005|
|Priority date||Jan 10, 2005|
|Also published as||EP1679613A2, EP1679613A3, US20060156226, US20110145701|
|Publication number||032817, 11032817, US 7937653 B2, US 7937653B2, US-B2-7937653, US7937653 B2, US7937653B2|
|Inventors||Hervé Déjean, Jean-Luc Meunier|
|Original Assignee||Xerox Corporation|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (14), Non-Patent Citations (24), Referenced by (8), Classifications (9), Legal Events (3)|
|External Links: USPTO, USPTO Assignment, Espacenet|
The following relates to the information storage and processing arts. It finds application in conjunction with electronic document format conversion and in particular with cataloging of legacy documents in a marked-up format such as extensible markup language (XML), standard generalized markup language (SGML), hypertext markup language (HTML), or the like, and will be described with particular reference thereto. However, it is to be appreciated that the following is amenable to other like applications.
Legacy document conversion relates to converting unstructured documents existing in formats such as Adobe portable document format (PDF), various text formats, various word processing formats, and the like into structured documents employing a markup language such as XML, SGML, HTML, and the like. In structured documents, content is organized into delineated sections such as document pages with suitable headers/footers and so forth. Alternatively, a textual header of a table spanning out multiple pages comprises a pagination construct when this header repeats itself on those pages. Such organization typically is implemented using markup tags. In some structured document formats such as XML, a document type definition (DTD) or similar document portion provides overall information about the document, such as an identification of the sections, and facilitates complex document structures such as nested sections.
There is thus interest in converting unstructured documents to a structured format when such structure can facilitate storage and access of this document as a “legacy document”. The particular motivations for converting documents are diverse, typically including intent to reuse or repurpose parts of the documents, desire for document uniformity across a database of information store, facilitating document searches, and so forth. Technical manuals, user manuals and other proprietary reference documents are common candidates for such legacy conversions.
A particular problematic issue that arises during the conversion process is the rebuilding or preserving of structural information. The output structure can be very different from the input structure. For example, page segmentation is often discarded in a logical representation, where logical units are elements such as chapters and sections (pages are usually considered as a physical element and do not appear). Content elements related to this page segmentation, typically headers and footers, and present in the input document have then to be processed cautiously. In prior art converters (PDF2[XML/HTML]), headers and footers are not differentiated from the body elements and can disrupt the flow of the main text. This not only generates an incorrect logical document, but can also introduce noise for further processing (natural language processing, for instance). Accordingly, the existing methods and apparatuses for identifying and extracting pagination constructs in the conversion of structured legacy documents is neither efficient nor robust. Of particular note is Xiaofan Lin, Header and Footer Extraction by Page-Association, Hewlett-Packard Technical Report, www.hpl.hp.com/techrepports/2002/HPL-2002-129.pdf, 2002. This reference relies upon comparison with neighboring pages for identifying a particular relationship indicative of commonly configured headers/footers. Such neighboring page comparison techniques can fail when the header/footer occurs very few times in the document.
For purposes of this application, “header” is intended to comprise matter (textual content) that is printed at the top of every page of the document typically positioned in the top margin of the page, for example, a title, page number, file name, revision dates, the author's name, or any other information about the document that is repeated throughout the document or a portion of the document. Likewise, a “footer” is similar information content positioned in the bottom margin of the page. As used in the subject application, “header/footer” should be construed to include either a header or footer individually or in combination.
The copending, commonly assigned applications comprise U.S. Patent Application Publication No. 20060155703A1, entitled, Method and Apparatus for Detecting a Table of Contents and Reference Determination, by Herve Dejean et al.; U.S. Patent Application Publication No. 06-0155700-A1, entitled, Method and Apparatus for Structuring Documents Based on Their Layout, Content and Collection, by Herve Dejean et al.; and U.S. Pat. No. 7,165,216, entitled, Systems and Methods for Converting Legacy and Projecting Documents Into Extended Markup Language Format, by Boris Chidlovskii, which are herein incorporated by reference.
The following provides improved apparatuses and methods that overcome the above-mentioned disadvantages and others to provide a light and robust method and system for detecting headers and footers of a document.
A method and apparatus is provided for detecting pagination constructs, i.e., textual elements of a document that were introduced by the pagination, which is particularly useful for more efficiently converting an unstructured document into a structured document during legacy document conversion. Pagination constructs can be characterized by the fact that their number of occurrences in the paginated document directly depends on the size of the pages in the document. At the extreme, each of the constructs would occur only once if the document were to fit on a single large enough page. Such constructs typically include the page headers and footers, but are not restricted to merely those types. For instance, it also includes a textual header of a table spanning multiple pages. When a document is segmented to pages, a model of the document can be assumed wherein the pages have a certain size to contain textual elements (text-boxes) with some coordinates (x, y, width, height) on the page and some textual content. By measuring text-box-to-text-box similarity as defined by such features as textual content, text-box position and area on a page, and the distance within the document in terms of number of pages the text-boxes are on, pagination constructs can be identified when the similarity is determined to be above a certain threshold.
A more particular method and apparatus is described for detecting a header/footer. In header and footer zones, it is a fact that the text content variability is much lower than in the main body of text content in a document page. The subject method and apparatus rely upon this factual basis to identify the header and footer zones in the document. Such methods and apparatus are especially useful as a preprocessing step in order to delete the noise introduced by headers and footers, and to better facilitate content-based processing, such as legacy document conversation. In one embodiment for identifying the header/footer of a document, a relative position of textual content of the document is identified. A textual variability of the content of the identified relative position is computed. The textual variability is compared with a predetermined textual variability known as indicative of a header/footer content. The header/footer of the document is then set by associating contiguous contents having the predetermined textual variability. The identifying of the relative position typically comprises allocating a relative vertical position, such as line numbering per page of the document. The textual variability is computed by fragmenting the text content into a number of text fragments with text blocks and differentiating the text fragments into predetermined different kinds. The ratio comprising the percentage of different kinds of text fragments and the total number of text blocks for a particular portion of the document is used as an indicator of the candidacy of that portion as a header/footer. A particular value of the relationship has been empirically determined to be a reliable comparative value.
In accordance with another aspect of the an illustrated embodiment, surrounding content elements to the header/footer lines identified as above are merged into the header/footer line list even where a textual variability computation indicates that such lines by themselves are not likely header/footer candidates, when the number of text blocks and the number of different kinds of fragments including text blocks in the surrounding lines is added to the pre-identified header/footer zone line list and there is no increase in the value of the textual variability for the overall list. Such a computation would generally indicate that the kinds of text blocks in the surrounding lines are already within the header/footer line list so that they are likely portions of the header/footer zone.
In accordance with another aspect of the disclosed embodiments, an apparatus is disclosed for identifying a pagination construct such as a header/footer in a document. A text fragmenter segregates the document into pages and differential text blocks on the pages. The text blocks are identifiable with respect to relative zones within the page, e.g., the top half or bottom half of the page. A processor computes a textual variability score (TVS) for the text blocks. The TVS comprises of the relationship between a total number of text blocks and a number of different kind of text blocks within a selected zone of the page. A selector selects a potential header/footer text contents comprising a compilation of text blocks having a TVS indicative of header/footer content whereby a selected text content is set as the header/footer. The selector first associates a header/footer content by page line computation of TVS for the textual content therein and secondly associates contiguous line content all having a preselected TVS indicative of header/footer context.
Various alternatives with the above-mentioned and further features and advantages will be apparent to those skilled in the art from the specific apparatus and its operation or methods described in the examples below, and the claims.
The subject embodiments have an overall purpose of detecting a zone at the top of a page of a document and at the bottom of a page of a document, which corresponds to a pagination construct such as a header or footer zone. Efficient characterization of such zones as shown herein, use the following heuristic: In a header or footer zone, the textual variety is much lower than in the body page of the document. The illustrated embodiments for header/footer detection may comprise at least four steps:
These steps, and others, will be described in detail with reference to the drawings and wherein
Text fragmenter 14 breaks the converted document into an order sequence of text fragments comprised of pages and text blocks 16, and which are typically loaded as a list of text strings from a text from an XML file produced from the document using an off-the-shelf converter. A paper document is suitably scanned using an optical scanner and processed by optical character recognition (OCR). For a text document, each line suitably becomes a fragment ordered line-by-line. For an XML or HML document, each PCDATA suitably becomes a text fragment. Fragmenter 14 also affects a position allocation of the document text content so that the page segmentation and vertical positions per page of the text content are preserved. For plain text format, the vertical position used is a line number. For proprietary format, the document can be converted into plain text format, or can be converted into a format where information about the vertical position exists. For example, a PDF2XML converter can provide page segmentation and positional information of the blocks of texts.
With reference to
With reference to
Once the text blocks with their positions have been identified, textual content is preferably normalized. All digits [0-9] are replaced by a unique character (D). Such normalization is preferably the only normalization applied and is due to the frequent presence of page/chapter/section numbers and date in headers/footers, which introduces a textual variability which is not relevant for the subject discussions.
A processor 18 effects the textual variability computation. The different positions of all the textual blocks in a document are listed. For each position a computation is performed for associating the number of text blocks occurring at the position and the total number of different types of fragments including text blocks occurring at the position taking into account the whole document, e.g. all the pages of the document. A textual variability score (TVS) for each position is computed as follows:
Table 1 below shows the fragments with text blocks occurring at a position 108 for all pages and their frequencies in the document (e.g., a technical manual).
Texts at position 108
Installing and Configuring 10/100Base-TX/9000
Configuring Network Connectivity Using SAM
10/100Base-TX Interface Card Statistics
Hardware Regulatory Statements
Hardware Reference Information
Table 1 shows seven different types of fragments with text blocks but the total number of all text blocks within all seven types is 93 text blocks. Applying Equation 1 to the text content of Table 1 would thereby result in a textual variability score at position 108 of 7/93 or 0.075.
Table 2 below shows a TVS score for six different positions in a page of the document.
#Different Text Blocks
The top element in the Table comprises the relative line position (e.g., 96, 108, etc.), the second line comprises the total number of text blocks at the line position (e.g., 5, 93, etc.), the third line number comprises the total number of fragments including text blocks at the position (e.g., 3, 7, etc.), and the last item for each set comprises the computed textual variability score (e.g., 0.6, 0.08, etc.).
As Table 2 shows, the textual variability of the element at position 108 has a very low score, while the textual variability of the element at position 156 has a very high score. The element 108 has 93 occurrences of text blocks (which means that 93 blocks of text were positioned at the location y=108). Among these 93 occurrences, only 7 were different kinds of text fragments. After the TVS for all of the lines of the document have been computed, they are stored in a database 20, 68 and based upon the scores, selective ones of the lines are identified as potential header/footer data and can be used for detecting the header/footer zone. More particularly, once the variability score for each content element (a content element is composed of a position number in the text blocks occurring at this position in the document) is computed. An effort is made to identify whether the document has header zones and footer zones. The method comprises identifying a first textual content element whose TVS is predetermined to likely comprise a part of the header/footer zone. The zone is extended from this textual element by merging all contiguous content elements in order to find the largest possible header/footer zone.
The identification number of potential header/footer textual content element is accomplished by preselecting a given TVS threshold which has been empirically determined to give a likely identification of a header/footer content element. In the present embodiment the threshold θ was set at 0.5 so that all content elements with a TVS lower than 0.5 are identified as potential header/footer textual content elements. The threshold can be selectively set at different values based upon empirical results for the type of document being converted for legacy storage and access.
Continuing with the example of the selective first top element and first bottom elements identified in Table 1, Table 3 below indicates which elements are potential header/footer textual content candidates:
#Different Text Blocks
Potential Candidate (θ = 0.5)
. . .
. . .
. . .
. . .
. . .
. . .
. . .
. . .
. . .
. . .
. . .
. . .
In order to define the top content element in the header zone and the bottom contact element in the footer zone, one starts from the top of the page for the header and at the bottom of the page for the footer. The first contact element starting from the top of the page or from the bottom of the page having a TVS indicative of a header/footer would be the initial textual content element of either the header/footer. For the header detection, the initial selected element must occur in the upper half of the page, and for the footer detection, the initial contact element must occur in the lower half of the page. If no element is identified that has the required TVS, then the subject method considers that the document has no header or no footer. In the example of Table 3, the element 108 is identified as the initial contact element and the element 684 as the initial footer content element.
Expansion of the header/footer zone beyond the initial contact elements comprises merging of contiguous elements to the initial content elements. More particularly, any contiguous textual content elements to the initial contact element having a TVS lower than θ also belong in the header/footer zone. Starting with the initial contact elements in the header and footer zones, a list with all the contiguous potential textual contact candidates is built. The augmented list comprises a new larger header/footer zone. A new TVS for the augmented list is computed 70 by applying the formula of equation 1 for the entire number of text blocks in the list as a divisor for the total number of different types of fragments including text in the list. (By construction, the new augmented list also has a variability score lower than θ). Continuing with the Table example, the element 122 is added to the initial element 108. The score of the header list is 0.23. The footer list is only composed of the element 684, since no adjacent element has a score below θ.
The merging process is especially useful in order to smooth the textual contacts elements position. In converted documents (for example, a PDF document converted in XML), a small variation in the position of elements per page can occur. For example, the position of the first paragraph in the body page can start at position 61 for some pages, 62 for other pages or 63. The subject method does not require a specific (pre)processing for these elements. The body page of textual contact elements thus corresponds to content elements contiguous to the header/footer zone whose line TVS is higher than θ.
It is another feature of the subject embodiments that surrounding line elements to the header/footer zone are further investigated for identifying previous and following elements whose variability is higher than θ, but which line elements should also be considered nevertheless as components of the headers/footers. The merging 72 of potential surrounding lines in processor 24 is affected by determination that the preceding and following elements are appropriate for insertion into the header/footer zone if a lines insertion decreases the TVS of the new augmented list. More particularly, and with references to
The steps for effecting merger of surrounding elements is to integrate textual content elements whose individual line TVS is higher than the threshold, but which can still be actual header/footer content elements.
Another aspect of the subject embodiments considers whether the identified list of header elements may not start from the top of the page. For example, with reference to
Alternatively, in order to reduce the processing time, only a selected number of the first n top elements, and a selected number of the first n bottom elements can be selected for processing in accordance with the steps identified above. As a practical matter, considering all elements of a document should not deteriorate the overall method performance, which should be very fast, no more than a few seconds for a 1,500 page document.
With reference to
For purposes of this embodiment, it is considered that the document is segregated 90 into a plurality of text-boxes. It is assumed that the model of the document, where the document is segmented into pages is comprised of pages having a certain size and containing textual elements with some coordinates (x, y, width, height) on the page and some textual content. It is a general model that most PDF2HML converters use. For purposes herein, such textual elements will be referred to as a “text-box”. The text-boxes are compared 92 for text-box similarity measure, taking values that take into account three features: first, their textual content, second, their positions in areas on the page, and third, their distance in terms of the number of pages they are on. More particularly, the similarity measure 92 is comprised of a combination of these three measures. First, textual similarity, is valued as 1 if the two texts are equal, or 0 otherwise. Alternatively, an edit distance, coupled or not with some text normalization can be used. Text normalization permits in particular the detection of a page number as a pagination element, while the strict equality test fails since each page is a different number. Second, geometric similarity may optionally comprise a page centering or text-box thickness. Page centering relates to the changing of the coordinate system of each page. The origin of the new coordinate system moves from the top left corner (typically) to the center of the box bounding all text-boxes on the page. The centering operation is useful in dealing with global translation of text-boxes between odd and even pages, which appears in dome documents, some other collections do not need it. Making the text-boxes border thicker by T pixels to prevent more displacements from one page to another is also useful in assessing geometric similarity where T is an empirically preselected value. The geometric similarity would then be defined 94 as the ratio of the surface of the intersection of two text-boxes by the surface of their union. The third measure for similarity comprises page proximity 96. It is assigned a value of 1 if the two text-boxes are on pages closer than P (a preselected number) of pages, 0 otherwise. This defines a window of page proximity of pages around each text-box out of which the similarity falls to 0.
Similar text-boxes can then be clustered 98 using a classical agglomerative single-link clustering algorithm. Data Clustering: a review, ACM Computing Surveys, (CSUR) Volume 31, Issue 3 (September 1999), A. K. Jain, M. N. Murty, P. J. Flynn. The similarity spatial data is set to 0.5. Clusters grouping elements having a similarity above θ are those containing pagination elements. Other clusters are ignored, including those with a single text-box inside.
The claims, are originally presented and as they may be amended, encompass variations, alternatives, modifications, improvements, equivalents, and substantial equivalents of the embodiments and teachings disclosed herein, including those that are presently unforeseen or unappreciated, and that, for example, may arise from applicants/patentees and others.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US5434962||May 24, 1994||Jul 18, 1995||Fuji Xerox Co., Ltd.||Method and system for automatically generating logical structures of electronic documents|
|US5491628||Dec 10, 1993||Feb 13, 1996||Xerox Corporation||Method and apparatus for document transformation based on attribute grammars and attribute couplings|
|US6298357||Jun 3, 1997||Oct 2, 2001||Adobe Systems Incorporated||Structure extraction on electronic documents|
|US6353840 *||Aug 13, 1998||Mar 5, 2002||Ricoh Company, Ltd.||User-defined search template for extracting information from documents|
|US6470306 *||Jun 9, 1997||Oct 22, 2002||Logovista Corporation||Automated translation of annotated text based on the determination of locations for inserting annotation tokens and linked ending, end-of-sentence or language tokens|
|US6487566||Oct 5, 1998||Nov 26, 2002||International Business Machines Corporation||Transforming documents using pattern matching and a replacement language|
|US6490603||Mar 19, 1999||Dec 3, 2002||Datapage Ireland Limited||Method and system for producing documents in a structured format|
|US7310773 *||Jan 13, 2003||Dec 18, 2007||Hewlett-Packard Development Company, L.P.||Removal of extraneous text from electronic documents|
|US20020143818||Mar 30, 2001||Oct 3, 2002||Roberts Elizabeth A.||System for generating a structured document|
|US20030093760||Nov 12, 2002||May 15, 2003||Ntt Docomo, Inc.||Document conversion system, document conversion method and computer readable recording medium storing document conversion program|
|US20030208502||May 3, 2002||Nov 6, 2003||Xiaofan Lin||Method for determining a logical structure of a document|
|US20040139384 *||Jan 13, 2003||Jul 15, 2004||Hewlett Packard Company||Removal of extraneous text from electronic documents|
|US20050154979 *||Jan 14, 2004||Jul 14, 2005||Xerox Corporation||Systems and methods for converting legacy and proprietary documents into extended mark-up language format|
|US20060101058 *||Nov 10, 2004||May 11, 2006||Xerox Corporation||System and method for transforming legacy documents into XML documents|
|1||Adler, S., et al., "Extensible stylesheet language (XSL), Version 1.0," W3C 2001, http://www.w3.org/TR/2001/REC-xsl-20011015/.|
|2||Aiello, M., Monz, C., Todoran, L., Worring, M., "Document understanding for a broad class of documents", International Journal on Document Analysis and Recognition (IJDAR), vol. 5, 2002, Springer-Verlag, pp. 1-16.|
|3||Anjewierden, A., "AIDAS: Incremental logical structure discovery in PDF documents", Proceedings of the International Conference on Document Analysis and Recognition (ICDAR), Seattle, 2001.|
|4||Anonymous, "Document Recognition and Retreval X (Proceedings of SPIE/IS&T): Header and Footer Extraction by Page-Association from Xiaofan Lin" Imaging Online Publications Catalog, [Online] vo. 5010, XP002543908, Retrieved from the Internet: URL:http://www.imaging.org/ScriptContent/store/physpub.cfm?seriesid=24&pubid=571 [retriefed on Jun. 23, 2009].|
|5||Belaïd, A., Pierron, L., Valverde, N., "Part-of-speech tagging for table of contents recognition", International Conference on Pattern Recognition (ICPR 2000), vol. 4, Sep. 3-8, 2000 Barcelona, Spain.|
|6||Dori, D., Doermann, D., Shin, C., Haralick, R., Phillips, I., Buchman, M., Ross, D., "The representation of document structure: A generic object-process analysis", Chapter 16, Handbook of Character Recognition and Document Image Analysis, World Scientific Publishing Company, 1997, pp. 421-456.|
|7||Dori, D., Doermann, D., Shin, C., Haralick, R., Phillips, I., Buchman, M., Ross, D., "The representation of document structure: A generic object-process analysis", Chapter XX, Handbook on Optical Character Recognition and Document Image Analysis, World Scientific Publishing Company. 1995/1996, pp. 000-000.|
|8||European Search Report, Application No. 06100200.2-1527/1679613, Dated Sep. 11, 2009, Berlin, Germany.|
|9||Herve Dejean, et al., "Versatile Page Numbering Analysis", Xerox Publications, [Online] XP002533713, Retrieved from the Internet: URL:http://www.xrce.xerox.com/Publications/Attachments/2007-031/2007-031.pdf [retrieved on Jun. 24, 2009].|
|10||Klink, S., Dengel, A., Kieninger, T., "Document structure analysis based on layout and textual features", Pcroceedings of Fourth IAPR International Workshop on Document Analysis Systems, DAS 2000, Rio de Janeiro, Brazil, 2000, pp. 99-111.|
|11||Lin, C.C., Niwa, Y., Narita, S., "Logical structure analysis of book document images using contents of information", 4th International Conference on Document Analysis and Recognition (ICDAR'97), Ulm, Germany, Aug. 1997, pp. 1048, 1054.|
|12||Lin, X., "Header and footer extraction by page-association", Hewlett-Packard Company Technical Report, 2002, www.hpl.hp.com/techreports/2002/hpl-2002-129.pdf.|
|13||Lin, X., "Text-mining based journal splitting", Proceedings of the Seventh International Conference on Document Analysis and Recognition (ICDAR 2003), vol. II, Aug. 3-6, 2003, Edinburgh, Scotland.|
|14||Lin, X., Simske, S., "Automatic document navigation for digital content re-mastering", SPIE COnference on Document Recognition and Retrieval XI, Jan. 18-22, 2004, San Jose, CA.|
|15||Power, R., Scott, D., Bouayad-Agha, N., "Document Structure", Computational Linguistics, vol. 29, No. 2, 2003, pp. 211-260.|
|16||Satoh, S., Takasu, A., Katsura, E., "An automated generation of electronic library based on document image understanding", Proceedings of the Third International Conference on Document Analysis and Recognition (ICDAR'95), vol. 1, Aug. 14-15, 1995, Tokyo, Japan, pp. 163-166.|
|17||Summers, K.M., "Automatic discovery of logical document structure", PhD thesis, Cornell University, Computer Science Department, Aug. 1998, pp. 1-181.|
|18||U.S. Appl. No. 11/032,814, filed Jan. 10, 2005, Herve Dejean et al.|
|19||U.S. Appl. No. 11/033,016, filed Jan. 10, 2005, Herve Dejean et al.|
|20||Virk, R., "Converting PDF files into XML", CambridgeDocs, 2004, www.cambridgedocs.com.|
|21||Xiaofan Lin, "Header and Footer Extraction by Page-Association" [Online] 2002, pp. 1-8, XP002533579, Retrieved from the Internet: URL:http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.11.6211 [retrieved on Jun. 23, 2009].|
|22||Xiaofan Lin, "Header and Footer Extraction by Page-Association" CITESEERX Abstract, [Online] 2002, XP002543906, Retrieved from the Internet: URL:http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.11.6211 [retrieved on Jun. 23, 2009].|
|23||Xiaofan Lin, "Header and Footer Extraction by Page-Association" Document Recognition and Retrieval X-DRR 2003: Santa Clara, California, USA (Layout Analysis), [Online] pp. 1-3, XP002543907, Universitat Trier, Retrieved from the Internet: URL:http://www.informatik.uni-trier.de/(ley/db/conf/drr/drr2003.html#Lin03 [retrieved on Jun. 23, 2009].|
|24||Xiaofan Lin, "Header and Footer Extraction by Page-Association" HP Labs Technical Reports, [Online] Dec. 26, 2002 XP002533592, Retrieved from the Internet: URL:http://web.archive.org/web/20021226015423/http://www.hpl.hp.com/techreports/2002/HPL-2002-129.html [retrieved on Jun. 23, 2009].|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US8478046||Nov 3, 2011||Jul 2, 2013||Xerox Corporation||Signature mark detection|
|US8645819||Jun 17, 2011||Feb 4, 2014||Xerox Corporation||Detection and extraction of elements constituting images in unstructured document files|
|US8812870||Oct 10, 2012||Aug 19, 2014||Xerox Corporation||Confidentiality preserving document analysis system and method|
|US8830487||Jul 9, 2012||Sep 9, 2014||Xerox Corporation||System and method for separating image and text in a document|
|US9008425||Jan 29, 2013||Apr 14, 2015||Xerox Corporation||Detection of numbered captions|
|US9008443||Jun 22, 2012||Apr 14, 2015||Xerox Corporation||System and method for identifying regular geometric structures in document pages|
|US9189461||Jul 16, 2013||Nov 17, 2015||Xerox Corporation||Page frame and page coordinate determination method and system based on sequential regularities|
|EP2765541A1||Jan 24, 2014||Aug 13, 2014||Xerox Corporation||Physical and electronic book reconciliation|
|U.S. Classification||715/234, 715/243, 715/239, 715/251|
|Cooperative Classification||G06F17/2745, G06F17/217|
|European Classification||G06F17/21F7, G06F17/27H|
|Jan 10, 2005||AS||Assignment|
Owner name: XEROX CORPORATION, CONNECTICUT
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:DEJEAN, HERVE;MEUNIER, JEAN-LUC;REEL/FRAME:016178/0359
Effective date: 20041213
|Jun 30, 2005||AS||Assignment|
Owner name: JP MORGAN CHASE BANK,TEXAS
Free format text: SECURITY AGREEMENT;ASSIGNOR:XEROX CORPORATION;REEL/FRAME:016761/0158
Effective date: 20030625
|Oct 16, 2014||FPAY||Fee payment|
Year of fee payment: 4